The contributions of anatomy and biophysical properties to the function of neurons and neural circuits are best understood with the aid of computer simulations. NEURON, a program which we have developed and provide freely for Mac OS X, MS Windows, and UNIX, has simplified the creation and analysis of neural models for nonspecialists in numerical methods and programming. It can simulate individual neurons and networks of neurons on workstations, clusters, and massively parallel supercomputers. Model properties may include, but are not limited to, complex branching morphology, multiple channel types, inhomogeneous channel distribution, ionic diffusion, extracellular fields, electronic instrumentation, and artificial spiking neurons. NEURON is used by neuroscientists around the world to investigate cellular and network mechanisms that are involved in inborn and acquired disorders such as epilepsy, multiple sclerosis, and disorders of learning and memory, and how they are affected by therapeutic interventions such as medications and deep brain stimulation. The community of NEURON users is growing steadily as increasing numbers of investigators, experimentalists and theoreticians alike, incorporate it into their research plans. Since neuroscience is a rapidly advancing field, investigators' needs are always changing. Consequently NEURON must continue to develop in several critical areas in order to continue to satisfy the evolving requirements of neuroscience research. There is a need for ever greater computational resources; we will address this by extending NEURON's new parallel features with selectable methods optimized for widely used classes of network topology. There is a need to access analysis tools that have become available in other biological sciences, the physical sciences, and engineering; to this end, NEURON is adopting a modern programming language (Python). There is a need for more flexible and powerful ways to perform dynamic clamp experiments; NEURON can be employed in this mode to great advantage, but the setup, hardware testing, and experiment configuration effort must be reduced. The NEURON community can fully exploit these advanced capabilities only if they are easy to install and use. We have also identified patterns of usage that call for the creation of new GUI tools to facilitate model specification: a finite state machine builder for specifying new synaptic mechanisms and artificial spiking cells, and a tool for specifying the geometry and kinetics of ion and second messenger accumulation mechanisms. These needs require concerted efforts in the areas of increasing performance, robustness, packaging, documentation, training, and user consultation and collaboration. In this proposal we present a research plan that is designed to address these areas.